what is xenodot-twin fr?

Xenodot Twin is a personal project that turns a real building or plant model into a live, data bound 3D digital twin inside the Godot game engine. It is built as an AI agent framework on top of the Claude Code SDK, and instead of a single chat prompt, it walks you through a structured pipeline of steps to produce a working 3D viewer. The process starts with an interview stage that pins down a small, buildable slice of the project and writes a one page design document. From there, a real building model in IFC or BIM format is converted into a Godot scene and joined to your data by element id. A separate step optimizes the scene so it runs smoothly, using techniques like level of detail and instancing to hit a target frame rate. Another step binds your master data and live time series values to the actual 3D elements, and a final verification step runs automated checks before anything is considered finished. The project is explicit about what it is not: it does not simulate physics, airflow, or structural behavior. It only shows what your existing 3D model looks like and what your existing data says, rendered in three dimensions with the ability to scrub back through recorded history. The README backs up its claims with measured numbers rather than vague promises. On one test machine, an optimization pass took a dense scene from about 27 frames per second to about 119 frames per second. Data joins between the 3D model and its source data reportedly succeed for all mesh elements, verified automatically rather than by hand. Playback of recorded data is checked using a cryptographic fingerprint to confirm two runs produce identical results. To try it, you need Claude Code, Node.js 18 or newer, and a Godot 4.x engine binary installed. After cloning the repository and running an install command, a guided setup script checks your environment, wires the project together, and verifies it boots correctly. A web interface then runs locally for you to view and manage the process. The project is released under the MIT license and currently has no stars, reflecting its status as a proof of concept rather than a finished product.